mysql - 如何组合 2 个 SQL 查询并检索累计计数？

Question

我正在尝试检索特定 A/B 测试组合每天的访问次数和转化次数。每个组合代表 A/B 测试的不同变体。在这里，我仅使用 '1' 和 '2' 来表示变体，但从技术上讲，可能会有更多变体。

我编写了以下 2 个独立工作的查询。是否可以组合这些或编写一个查询来检索我想要的数据？

访问查询:

SELECT DATE(visit.created), visit.combination, COUNT(visit.id) 
FROM visit 
WHERE visit.user_id = 6
AND visit.experiment_id = 1
GROUP BY DATE(visit.created), visit.combination

访问结果:

转化查询:

SELECT DATE(conversion.created), conversion.combination, COUNT(conversion.id) 
FROM conversion 
WHERE conversion.user_id = 6
AND conversion.experiment_id = 1
AND conversion.goal_id = 1
GROUP BY DATE(conversion.created), conversion.combination

转化结果:

另外，如果我能检索到运行总(累积)计数，那就太好了，如下所示，请参阅最后两列。我按组合对下表进行了分组，这样累计计数更容易理解:

+---------------+-------------+----------------------+-----------------+--------------+--------------+
| DATE(created) | combination | COUNT(conversion.id) | COUNT(visit.id) | cumulative_c | cumulative_v |
+---------------+-------------+----------------------+-----------------+--------------+--------------+
| 2015-11-17    | 1           |                    1 |               3 |            1 |            3 |
| 2015-11-18    | 1           |                    7 |               4 |            8 |            7 |
| 2015-11-19    | 1           |                    3 |               8 |           11 |           15 |
| 2015-11-17    | 2           |                    4 |               1 |            4 |            1 |
| 2015-11-18    | 2           |                    2 |               6 |            6 |            7 |
| 2015-11-19    | 2           |                    9 |               6 |           15 |           13 |
+---------------+-------------+----------------------+-----------------+--------------+--------------+

数据库模式:

Answer 1

合并非常简单:添加 0 值列，执行 UNION_ALL，然后再次分组并求和。

SELECT dt, combination, SUM(v_count) as v_count, SUM(c_count) as c_count
FROM 
(
SELECT DATE(visit.created) as dt, visit.combination as combination, COUNT(visit.id) as v_count, 0 as c_count
FROM visit 
WHERE visit.user_id = 6
AND visit.experiment_id = 1
GROUP BY DATE(visit.created), visit.combination

UNION ALL

SELECT DATE(conversion.created) as dt, conversion.combination as combination, 0 as v_count, COUNT(conversion.id) as c_count
FROM conversion 
WHERE conversion.user_id = 6
AND conversion.experiment_id = 1
AND conversion.goal_id = 1
GROUP BY DATE(conversion.created), conversion.combination
) as t
GROUP BY dt, combination

现在，计算总计。在更高级的 DBMS 中，这称为“窗口”或“分析”函数。例如，在 Oracle 中你可以这样做:

SELECT dt, combination, SUM(v_count) OVER (PARTITION BY combination ORDER BY dt) as v_cumulative

对于上面的查询，它会给你你想要的。但是，MySQL 没有这样的功能。有方法绕过，描述here和 here例如，但它们非常棘手。